On Thu, 2023-03-16 at 09:36 +0100, Alexander Kanavin wrote: > On Thu, 16 Mar 2023 at 09:30, Richard Purdie > <[email protected]> wrote: > > In the openssl source this is generated by a perl script, > > rypto/modes/asm/aes-gcm-avx512.pl > > > > Wearing suitable safety equipment to stop it burning, inside the perl > > script you can see references like: > > > > my $rndsuffix = &random_string(); > > > > I think this is done to avoid symbol clashes but I couldn't see > > anything that would actually clash so more understanding of that this > > is doing is needed. I wondered if we could set some kind of perl > > seeding to make it deterministic. > > > > I'll reply again if I learn anything more. > > Even if it's not a real fix, avx-512 support can be more or less > disabled by default, as support for that is limited to high end server > grade chips. We haven't enabled avx/avx2 yet :)
I think the fix might be straight forward and just to add something like: # Ensure the local labels are reproduicble srand(10000); I'll put that into testing, see how it looks. Cheers, Richard
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#178685): https://lists.openembedded.org/g/openembedded-core/message/178685 Mute This Topic: https://lists.openembedded.org/mt/97621752/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
